Fuel Economy Variability Investigations: From Test Cells to Real World 2017-01-0894
Improving fuel economy has been a key focus across the automotive industry for several years if not decades. For heavy duty commercial vehicles, the benefits from minor gains in fuel economy can lead to significant savings for fleets as well as owners and operators. Additionally, the regulations require vehicles to meet certain GHG standards which closely translate to vehicle fuel economy. For current state of the art fuel economy technologies, incremental gains are so miniscule that measurements on the vehicle are inadequate to quantify the benefits. Engineers are challenged with high level of variability to make informed decisions. In such cases, highly controlled tests on Engine and Powertrain dynamometers are used, however, there is an associated variability even with these tests due to factors such as part to part differences, deterioration, fuel blends and quality, dyno control capabilities and so on. This variability grows dramatically during controlled vehicle track testing and eventually in customer trucks where driver habits and environmental factors can contribute significantly to the final realized fuel economy. Although, this information is intuitive, little research literature exists that attempts to quantify it.
This paper describes the variability seen in test cells, powertrain dyno, test tracks and real world driving by statistically analyzing data from past several years. The contribution from factors such as idle times, fuel quality, ambient conditions and road friction changes are further quantified using vehicle level simulations and other modeling techniques. Statistical methods to analyze and compare data at various stages of product development are discussed. The end customer should find this information useful to determine if the observed fuel economy discrepancies are controllable or uncontrollable and take corrective actions where possible. Policy makers will be able to use it to assess how the regulatory levels pan out in real world operating conditions.